The quote “all models are wrong, but some are useful” by Box and Draper  reflects today’s discussion about climate modelling and how much people think we can extract from them: If we desist from wanting to use models to depict “the true future”, which they cannot, a lot can be learned from using climate models.
A model generally is a representation or imitation of the real world based on the knowledge of the creator of the model. In case of climate models, they include climate system components (like the atmosphere, land surface, ocean and sea ice) governed by physical and biogeochemical processes, which are translated into mathematical equations, and then numerically solved on a computer. Fundamentally, a climate model is used to learn more about each component and its influence on the climate system and, with that, the exploration of past causes of a change in climate and simulation of the future climate.
Over the last years, there have been great advances in climate modelling, but this diversity of models has also made it easy to get lost in this model jungle. Generally, models can be distinguished following a hierarchy based on the model’s complexity (i.e. level of detail) and comprehensiveness (i.e. number of climate components) (for an overview see fig. 1) . This does not mean that the most complex model is the universally “best” model . Also, different model types cannot substitute each other. They rather complement each other dependent on the research or policy question, the need for temporal, spatial resolution or level of complexity. But how can we, who are not climate modellers make sense of them? I think that if you are generally interested in learning more about climate change, it is important to know some fundamental models. So I have tried to shed some light into this very dense jungle.
I divide between Energy Balance Models (EBMs), General Circulation Models (GCMs) and Earth System Models (ESMs). There are more models than that, but these cover the most general ones. While you read, you can come back to this picture below (fig 1), as it gives you a help for navigation between the model types.
Simple climate models
Energy balance models
One of the oldest and most basic models discussed here are energy balance models (EBMs), which include only a few parameters. The most common EBMs consist of the values of natural (e.g. solar) and anthropogenic (i.e. human made) forcing and some constants, which leads to an output of temperature and associated heat fluxes . Even though the input of the model is restricted to only a few parameters, they can give important insight on climate sensitivity or the climate on other planets . As they include only a few parameters, by themselves they cannot answer any more complex questions, have poor spatial resolution, and are mostly one or zero dimensional. This simplicity however, makes them quick to run.
To include more dimensions, an expansion of the simple EBM can be so called box diffusion models, which have more dimensions and deal with interactions of different systems (such as ocean, land, and atmosphere) represented as boxes interconnected through energy transfer equations. With each system (box) containing their own independent characteristics e.g. diffusion of heat into ocean layers . One of those simple climate models (SCMs), known as MAGICC (Model for the Assessment of Greenhouse-gas Induced Climate Change) has a very important role for the Intergovernmental Panel on Climate Change (IPCC) in terms of model intercomparison. This box model can be calibrated with parameters of more comprehensive climate model’s parameters to emulate their outputs.
Box models still parameterise many events, and do not permit very detailed questions related to climatic mechanisms, which can be considered a disadvantage. Their advantage over the more complex models is their very low computational effort. That way, they can do multiple runs, tuned with more complex models’ values, in much less time than those models themselves could.
Complex Climate models
General Circulation Models
General Circulation Models (GCMs) are much more comprehensive than EBMs, including more physical constituents, which are also represented in a more complex manner. In contrast to EBMs, they don’t conceptualise the world as a dot, a line or parts of it as boxes interacting, but incorporate the three dimensional nature of ocean and atmosphere . These 3-dimensional grid boxes have become smaller and more numerous. This fine resolution and the high diversity of processes represented in the grid boxes allows them the models to independently simulate circulation patterns of the ocean or the atmosphere . Besides a high spatial resolution, their high temporal resolution helps them perform seasonal or decadal projections . This helps them in simulating past and future climates to determine which cycles are the most important factor for climate response in general .
GCMs started off being either an oceanic or an atmospheric circulation model, which later got combined to form the Atmospheric Ocean General Circulation Models (AOGCMs) . The more complex the system, the more computational time and power is required for the model to run all integrations. Further adding components has made GCMs able to deal with a variety of problems. However, that way, they have become more complex and difficult to understand . Figure 2 shows add-ons throughout the last 50 years.Earth System Models
From the GCM models “upwards”, complex models mostly include a 3D geographical structure. The most complex and comprehensive models in the hierarchy today are Earth System Models (ESMs), which can be considered an expansion on general circulation models . They contain all of the present day’s knowledge, not only about the climate system, but also about other components and feedbacks. For example, they contain interactive representations of many biogeochemical cycles (Carbon, Sulfur, Ozone) . These models are so complex, a single run may take up to weeks, even months  and only a small range of situations can be explored in a few runs. Not many places in the world can perform this state of the art high-performance computing leading to a limited number of research groups in that area. For further model development, it is necessary to consider not only computationally expensiveness, but other limitations, even power demands and heat generated during the simulation runs . In principle, they can perform experiments intended to assessing various climate change mitigation strategies (forest/land use, geo engineering, ocean fertilisation) , so that we don’t have to try out ideas that might have a fatal impact but instead see what happens if we apply them to the models. For example, using AOGCMs can provide future projections on the outcome of us emitting greenhouse gases or aerosols .
But that might still leave us with the question of why the model outputs from these very complex models looks the way it does. They have become so complex that even their creators cannot understand them anymore. There is help from again, being more simple or at least “intermediately complex”.
Models of Intermediate Complexity
They include all relevant components of the earth system in sufficient detail, and have the component of the research question modelled in more detail. That makes them computationally efficient and easier to analyse than GCMs . Also, they can be used for simulating long term climate changes of the past to understand underlying climate dynamics in ESMs . However, the different model structures, assumptions, and choice of components that are represented in more detail, make model results not always reproducible by other models . But reproducibility of results is important not only in climate science, but ubiquitously across disciplines.
Climate models for policy makers
In addition to using climate models to answer scientific questions, they also serve to support in decision making for policy makers. These models can be used to consult in the application of certain adaptation and mitigation strategies and need to deliver a low level of uncertainty, and a high spatial resolution . Ideally, ESMs or GCMs should be used to simulate local impacts of climate change. But data coverage, calculation time needed and the way certain physical processes are described mathematically in the models prevent the complex models to scale down that much. There is yet a “scale gap” of about 30 km between global climate projections and a necessary resolution for impact studies . To close the scale gap, for example regional climate models are “nested” into the global models . This means that similar to EBMs, regional climate models use data generated from GCMs for their simulations. Other approaches are possible, where the application of different models or different physical descriptions is involved. I will not go into further detail here since this is just a quick overview.
With our current climate models, we managed to model something as complex as our planet’s climate with many of its interactions. We have created powerful tools to look into the past or the future, test our hypothesis and find out what impact human emissions have on the planet. But we are not done yet and it is more important than ever to strive for improving the performance and future credibility of the models ( as urged in ). If we want to make the most use of the present models and build our future models upon them, creating a hierarchy or ancestral chart (like in Figure 2) would be an important step in creating a map out of the climate model jungle.
Wanna keep digging?
I am by no means an expert in this area but wanted to share my personal findings with you to get a little overview over things. So if you want to learn more yourselves, I really recommend the book “A Climate Modelling Primer”, by Kendal McGuffie, Ann Henderson-Sellers . It is very entertaining and well written, explains the basics and applications for climate models. The examples are easy to grasp and they refer to a lot of interactive online sources.
Here are some links to EBMs you can play around with yourself:
MAGICC is open-source and if you are curious to lean more, it gives you an excellent opportunity to explore the future climate and its uncertainties. Its simulations are based on the results of more complex model’s simulations. If you wonder how climate change might influence the region you live in in the future, have a play around with this model. It can be downloaded from this website: http://www.cgd.ucar.edu/cas/wigley/magicc/ .
1. Box, G.E.P. and N.R. Draper, Empirical model-building and response surfaces, p.424. 1987: Wiley.
2. Houghton John T. , M.F.L.G., Giggs, David J. , Maskell, Kathy An Introduction to Simple Climate Models
used in the IPCC Second Assessment Report. IPCC Technical Paper II, 1997.
3. Flato, G.M., Earth system models: An overview. Wiley Interdisciplinary Reviews: Climate Change, 2011. 2(6): p. 783-800.
4. McGuffie, K. and A. Henderson-Sellers, Climate modelling primer Kendal McGuffie, Ann Henderson-Sellers. 2014: West Sussex:Wiley-Blackwell, 2014.
5. Flato, G., et al., Evaluation of Climate Models, in Climate Change 2013: The Physical Sci ence Basis. Contribution of Working Group I to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change, T.F. Stocker, et al., Editors. 2013, Cam bridge University Press: Cambridge, United Kingdom and New York, NY, USA. p. 741–866.
6. Weber, S.L., The utility of Earth system Models of Intermediate Complexity (EMICs). Wiley Interdisciplinary Reviews-Climate Change, 2010. 1(2): p. 243-252.
7. Hawkins, E. and R. Sutton, The potential to narrow uncertainty in projections of re gional precipitation change. Climate Dynamics, 2011. 37(1/2): p. 407.
8. Provenzale, A., Climate models. Rendiconti Lincei: Scienze Fisiche e Naturali, 2014. 25(1): p. 49.
9. Jakob, C., Going back to basics. Nature Clim. Change, 2014. 4(12): p. 1042-1045.